Reproduce “GTA: Generative Tower of Agents for Scalable Holonic Scene Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zeng Tao, Yunuo Chen, Xiang Feng, Yumeng He, Peihao Li, Nanxiang Jiang, Yaowei Guo, Fanlian Zeng and 2 more
GTA builds 3D scenes through holons, units that are both construction tasks and parts of larger scenes. Agents plan these units recursively, generate leaf assets, and assemble groups bottom-up, with independent work dispatched in parallel. In general scenes, 8.54% of objects exceed 1 mm penetration, versus 15.29% for the best baseline; generation is faster, but tests cover 40 scenes and up to 1,000 objects.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.